Tutorial｜B.AI Skills: Minimalist Deployment and Operation Guide, Injecting Financial Capabilities into AI Agents

Today, as large language model technology iterates at breakneck speed, artificial intelligence is rapidly restructuring productivity across countless industries. However, amid this exciting progression, the industry consistently faces a critical bottleneck in real-world deployment: most current AI Agents remain stuck at the level of "information processing" and "dialogue interaction." When it comes to rigorous fund transfers, asset management, or complex financial operations, they are often constrained by the lack of professional execution interfaces, trapped in a predicament of "understanding requests but being unable to take action."

To bridge the gap from "cognition" to "execution," AI urgently needs a dedicated underlying financial operating system. This is precisely the original vision behind B.AI. As a pioneer focused on building AI financial infrastructure, B.AI's core mission is to lay down native financial service systems and underlying communication pathways for the countless AI Agents of the future. To completely break through this execution barrier, the B.AI team has deeply encapsulated complex financial logic atop common industry-standard agent architectures and open-sourced the Skills module dedicated to AI Agent financial operations.

This article will guide you through the Skills offered by B.AI in an accessible way, showing how they empower AI Agents with genuine financial execution capabilities, and provide step-by-step, hands-on guidance for local deployment and practical invocation.

1. Deconstructing Modular Financial Skills Exclusively for AI Agents

In simple terms, the AI model itself possesses powerful text processing and logical generation capabilities, much like a "brain" with vast knowledge but lacking limbs. Skills are the "hands, feet, and professional toolkit" bestowed upon this brain. Today, leveraging modular Skills to expand AI's execution power has become the standard across mainstream agent frameworks. It encompasses a series of reusable, task-specific functional instruction sets.

Each Skill provided by B.AI encapsulates highly specialized financial and data domain knowledge and offers a highly automated, unified deployment solution. Through minimal Skill deployment, an AI Agent can execute cross-platform complex financial operations, payment protocols, and account asset management based on natural language instructions.

Its core value is reflected in:

• Expanding Capability Boundaries: Transforms AI from a mere "chat assistant" into a true "intelligent operator" with financial execution capability.
• Deterministic Execution: Constrains the AI's behavioral path, ensuring each transaction and query adheres to financial business security standards, thereby significantly reducing risks associated with AI hallucinations.
• Lego-like Composability: Focuses on providing high-value, general-purpose financial utility Skills. These not only seamlessly integrate with multiple mainstream agent frameworks but also allow different Skills to cooperate, forming complex financial workflows, thus truly solidifying the underlying connection between the global AI network and financial infrastructure.

2. Skills Installation Tutorial

Thanks to the unified installer provided by B.AI, even users without development experience can quickly complete end-to-end local deployment, enabling the AI Agent to leap from "general conversation" to "professional financial execution" at the press of a button.

1. Environment Setup

Before starting the installation, please ensure your local development environment is ready:

• OpenClaw or other compatible AI agent runtime environment.
• Node.js: Required on all platforms for one-click quick installation via npx.
• Python 3 & Git: Used for script processing and code cloning.

Tips: After running the installation script, the system will automatically perform checks, and users can configure settings according to the system prompts.

2. Run the Installation Command

Open the corresponding command-line tool according to your operating system (Terminal for macOS/Linux users, Git Bash for Windows users), copy and input the following command, and press Enter: npx skills add https://github.com/BofAI/skills.git

The system will prompt you to install the skills@1.4.9 (version number may update) base package. Input y to confirm and continue, after which the system will automatically clone the repository.

3. Select Skills as Needed

Once cloning is complete, the terminal will automatically scan and list the 8 currently supported core Skills. You can browse them using the up and down arrow keys, select or deselect items by pressing the spacebar, and confirm your selection by pressing Enter. The Skills currently offered by B.AI include:

• SunSwap DEX Trading: Supports quoting, token swapping, liquidity and pool operations.
• SunPerp Perpetual Futures Trading: Perpetual contract market data queries, order placement, and position management.
• TRC20 Token Toolkit: General token metadata queries, balance checks, and transfers.
• TronScan Data Lookup: Comprehensive blockchain data, transactions, tokens, and network-wide statistical queries.
• Multi-Sig & Account Permissions: Native multi-signature workflows and account permission management.
• recharge-skill & x402-payment: Account top-up and professional paid API resource protocol.

4. Select Installation Scope

After selecting the Skills, the system will ask for the installation scope:

• Project (Recommended): Installed in the current project's directory (e.g., ~\.agents\skills\), convenient for committing and versioning along with your project code.
• Global: Installed globally, effective for all local projects.

After making the selection, the installer will automatically complete the subsequent file linking and configuration. When "Done" is displayed at the bottom of the terminal, it indicates that the Skills have been successfully installed.

3. Practical Interaction Guide Using OpenClaw

Once the installation command is complete, your AI Agent (taking OpenClaw as an example) will have formally loaded this professional set of financial operation Skills. Next, you can command your AI Agent in the OpenClaw interface using the three typical scenarios below.

Scenario 1: Query Currently Loaded Skills (Status Confirmation)

Before starting complex tasks, if you want to confirm whether the tools you just installed are active, you can simply ask the AI in natural language to "take inventory." For example, "Please check and list the Skills currently loaded in your local directory." OpenClaw will automatically check and list the deployed core modules, such as sunswap-dex-trading, agent-wallet, tronscan-data-lookup, etc.

Scenario 2: Implicit Trigger (Natural Language Driven Execution)

This is the most intuitive interaction method. Users don't need to mention specific underlying code files; they just state their final request directly, like instructing a human business assistant. For example, "Please check the current real-time exchange rate for swapping USDT to TRX on SunSwap." Upon receiving the natural language command, OpenClaw will precisely match and activate the relevant Skill in the background. It will then automatically call the corresponding interface to complete the query and report the final result directly to you.

Scenario 3: Combined Invocation (Lego-like Coherent Workflow)

Thanks to the "Lego" characteristic where Skills are independent yet seamlessly combinable, users can give the AI compound tasks involving multiple steps, allowing it to autonomously invoke different components to complete the loop. For example, "First, check the current USDT balance in my address (enter your address). Then, using the best current exchange rate on SunSwap, estimate how much USDT I could get if I swapped all my TRX."

Special Note:

In the scenarios above, we have demonstrated the comprehensive capabilities of B.AI Skills. However, please note: If an instruction given to the AI Agent involves actual fund transfers, asset swaps, or substantive on-chain transactions like multi-signature authorization, it requires the use of B.AI's specially developed Agent Wallet component. Agent Wallet is a secure signing environment designed specifically for intelligent agents. When AI executes financial tasks, it uses an innovative dual-lock mechanism ("encrypted file + authorization password") to securely isolate the AI Agent's runtime environment.

To explore more underlying architectures and cutting-edge features that empower AI Agents, please visit the official B.AI website (https://b.ai/) for an in-depth experience and join B.AI in ushering in a new era of the agent economy.